Neural Techniques for Source Identification

Noise source separation is a key issue in environmental noise assessment and of particular interest in the implementation of the European Environmental Noise Directive (END), since the contribution to the overall noise level from each single source should be evaluated separately. This, because concerning noise reduction measures each noise source should be eventually reduced independently from the other sources and the effect of the single noise source reduction should be readily compared to the corresponding noise of other sources, and to the health benefits. A technique which is based on wavelet analyses was tested to evaluate how far these concepts could be developed to effectively assist reaching these objectives. This technique is applied here and evaluated using noise long-term measurement data of campaigns performed in Italy in the context of the HARMONOISE and IMAGINE projects funded by the European Commission. Moreover, a proposal is made for using the same techniques to assess the physiological human response to specific noise sources. A metric is introduced which considers each single event in relation to the environment where this happens and to the subjective feeling that this might evoke in the person. Brief time periods of the noise recordings obtained during the experimental campaigns in a real urban environment were used to test the technique. Subsequently, an attempt was done to separate the noise of the cars, motorbikes, buses, airplanes, trains and that produced by the local human voices. Characteristics of the sound signal will be shown, which are used to discern a specific sound source signal from another; these could be used in the future as alternative noise indicators.

MC LACHLAN Neil;  PAVIOTTI Marco;  KEPHALOPOULOS Stylianos;  KANT KUMAR Dinesh; 

2008-02-28

International Insitute of Acoustics and Vibration (IIAV)

JRC37746

www.icsc14.com,    https://publications.jrc.ec.europa.eu/repository/handle/JRC37746,